Indtroduction


I will be analysing the jazz standard Ornithology, first recorded by Charlie Parker. Ornithology is a refference to Parkers nickname “Bird”. Jazz standards often differ from one another quite alot, yet Ornithology is one among them that is particularly well known for all the different recordings. One reason for this is that the jazz standard has also been recorded under another name “How high the Moon”, which follows the exact same chord progression as Ornithology, but is usually in a slower tempo and follows a somewhat different melody. In essence though both ‘titles’ are the exact same jazz piece

I want to know if there are any real measurable differences in recordings under the name How High the moon and Ornithology. What makes some musicians name the recording how they do. What are some key aspects that differ causing the different name, or is this an arbitrary prefference of the musicians, and are the names really interchangable. In order to get Find out if there are any differences I created a playlist containing ±100 How High The Moon recordings, and ±100 Ornithology recordings.

Visualisation of Tempo in ‘Ornithology’ and ’How High the moon;


One of the first distiguishable features might be the tempo. As I stated earlier I believed that How high the Moon (HHTM) recordings usually were recorded in a slower tempo than Ornithology (OTHGY) recordings. However, this is not backed up by my findings.

On the left you will see a histogram the temp of all the recordings I added to my playlist. As You can see there are acctually no real differences in tempo between the recordings. Most recodings seem to cluster around 120 bpm. Yet my statement might not be completely untrue, as there are, just marginally, more slow HHTM recordings.

Popularity vs Tempo.


It might be that the dataset I created, happened to only include unusually fast HHTM recordings. Iff this were the case, then the popularity of the faster HHTM tracks must be less then the popularity of the, more usual, slower recordings of HHTM. Here we see the popularity versus the tempo for both titles.

This graph shows us that my dataset does not consist of outlier recordings of HHTM, as the popularity of all different tempos of HHTM recordings relates well to the same tempo OTHGY recordings. Further more this graph tells us that there is a definite sweet spot in tempo, where the recordings are more popular. This occurs for both How High The Moon and Ornithology around 115BPM. At a slower tempo of 80BPM another spike in popularity is seen, telling us that this song, in both titles, is also enjoyed at a slower tempo. What is interesting to see is that eve though the original recording was titled Ornithology, the How high the Moon recordings seem to be more popular overal. Another thing we can observe, is that my dataset actuallually contains a couple faster HHTM tracks that are much faster then any OTHGY tracks. I don’t per see see these as outliers, yet I do think that there are OTHGY recordings out there, that at least equal the tempo. Next up, how do two recordings of HHTM and OTHGY relate to one another structure wise.

similarity matrix and relation between hhtm and ornithology.


As you can see these recordings dont seem to be very similar pitch wise according to this similarity matrix…

Even though both tracks are in the same key, and folow the same chord scheme there does not seem to be a lot of similarity. An explenation for this is improvisation. Both pieces might be folowing the exact same scheme, but the musicians improvisation appears to make the recordings inheritly different. At least so viewing from a similarity matrix. We can try and view both structures by making selfsimilarity matrices.

Ornithology: self-similarity matrices for pitch and chroma.


In these two matrices you can see that there is definately a structure within the complex improvisation of OTHGY. It is not per see any recurring patterns which make me conclude this - the whole point of improvisation being that you play around with tempo, pitch and timbre - but it is the clear ‘new’ unplayed parts. The vertical yellow lines (chroma) mean that from there on a new part of the standard is played. So even though the musician is doing all kinds of complicated improvisation, the structure of the standard can be followed. If I am correct, the recordings of high how the moon should show somewhat similar paterns. However, we might see some difference, which might point to why some recordings are names Ornithology, and others How High the moon.

How high the moon: self-similarity matrices for pitch and chroma.


In the chroma matrix for HHTM we can indeed see a couple of vertical yellow lines that line up with that of OTHGY. Though it also seems to be that there is less imporovising in this recording (less yellow overall). In the Timbre matrix we see an interesting pattern. It seems like the recordings starts out with a clear(ish) theme, then does a lot of improvisation and ends with the theme again. The second time the theme is played, it is not a complete duplicate of the first time, it does still seem to have some nice improvisational quality to it. A structure like this seems very plausible to me in jazz. This however is not a structure we see in any of the other matrices (chroma and timbre) of this recording and the OTHGY recording. This might actually be one of the crucial differences between the two titles.

scatterplot energy v valence


From the previous similarity matrices, it is hard to make hard conclusions about what makes one recording be named so, and another different. So lets try something a bit different. On the left we see a scatterplot of valence vs energy, ere size represents the popularity. What we can see from this, is that for a large part the songs overlap completely. Yet in the left bottom corner, with low energy and low valence we see only HHTM recordings. Finally an aspect in which the titles really do differ. One might also argue that top left, or low valence high energy, is primarily occupied by Ornithology. Yet such an statement is not possible with the dataset we have. You would most likely need a much larger dataset is you would want to back up such a claim. Knowing however that there are some differences like that. We can try and see if we can use machine learning to create a classieifer which will identify a title, given a recording. Lets give it a go.

classification


The classification actually performes alot better then I would have expected. You can see that the wrong predictions are much outweighed by the accurate ones. This means that, as we have come to see already, there are some distinguishable features, but that there is also some overlap.

Conclusion

My very first hypothesis that it was tempo, that dictated the title, was quickly seen to be untrue.

It seems to be that the main difference in the recordings is that there is more improvisation in recordings titled OTHGY then in recordings titles HHTM.

Apendix

Here is all the information about all the used songs

playlist: ornithology <- get_playlist_audio_features(“1115668083”,“3pMfdAOvErb9AEnAO9W2fa”)

popular OTHGY: orni_maxPop <- get_tidy_audio_analysis(‘6A1mh31895sUIYEpOfeqC3’)

popular HHTM: hhtm_maxPop <- get_tidy_audio_analysis(‘5ldNRCiBHlFlBlvJmDMlCn’)